3D Multi-poses Face Expression Recognition Based on Action Units
Bai Jingxin, Yinan Lu, Shuo Zhang · 2019
This paper presents a new 3D multi-poses facial expression recognition method based on action units. Firstly, automatic landmark detection is carried out according to the detected position of nose tips. Then a new feature descriptor is created to describe facial features involving geometric, topological and texture information. AUs are detected by classifying the fused features, and further used for recognizing the facial expressions. The method achieves good performance on Bosphorus database.